disease_differential_expression.utf8.md

Data analysis

This analysis is concerned with identifying differentially expressed genes across disease states. Primarily we are interested in whether there is a particular gene expression signature that is associated with PSC/UC as opposed to UC alone. There are some complicating factors in the analysis in that a small subset of samples are mildly inflamed - this could affect differential expression results. Nevertheless, I have presiously shown that there are no statistically significant differences betwen PSC/UC and UC in terms of Nancy scores at each of the tissue sites sampled.

Dataset at a glance

I have decided to perform differential expression analysis at each site separately and then compare the results. Initially I am using the LRT test in DESeq2 to define differentially expressed genes between healthy, PSC/UC and UC. The idea is that we could do something similar to the between tissue analyses if the differences in disease groups are not too subtle. First I will perform some further exploratory analysis.

I will base some of the filtering steps on the number of samples that I have in each tissue/disease group. Below is a summary of these numbers. The filtering is the same as was performed for defining differentially expressed genes between tissues. This allows a more straightforward comparison between analyses.

	HEALTHY	PSC/UC	UC
Caecum	9	7	10
Ileum	11	7	9
Rectum	12	7	10

Description of the metadata

Significance of metadata variables associated with group status is explored here. I feel like these are better as a plot rather than a table but could be wrong. Plot age, sex, nancy.score.at.site, UCEIS,

## Sex:
## Age:
## nancy:GI#2876-HEALTHY
##  nancy:GI#2902-HEALTHY
##  nancy:GI#2907-HEALTHY
##  nancy:GI#2911-HEALTHY
##  nancy:GI#3180-HEALTHY
##  nancy:GI#3189-HEALTHY
##  nancy:GI#3352-HEALTHY
##  nancy:GI#3859-HEALTHY
##  nancy:GI#3938-HEALTHY
##  nancy:GI#3951-HEALTHY
##  nancy:GI#4134-HEALTHY
##  nancy:GI#4221-HEALTHY
## UCEIS:GI#2876-HEALTHY
##  UCEIS:GI#2902-HEALTHY
##  UCEIS:GI#2907-HEALTHY
##  UCEIS:GI#2911-HEALTHY
##  UCEIS:GI#3180-HEALTHY
##  UCEIS:GI#3189-HEALTHY
##  UCEIS:GI#3352-HEALTHY
##  UCEIS:GI#3859-HEALTHY
##  UCEIS:GI#3938-HEALTHY
##  UCEIS:GI#3951-HEALTHY
##  UCEIS:GI#4134-HEALTHY
##  UCEIS:GI#4221-HEALTHY
## Urso:
## vedolizumab:
## AZA.6MP.ISS:
## X5.ASA:
## steroid:

Principle components analysis

This is an initial look at principle components using log2cpm. I have already had a look at this as an exploratory analysis and removed patients that were outliers i.e. their expression profiles were inconsistent with tissue patterns and may have been the result of a sample swap. Also removed was a patient that had polyps.

PERMANOVA assessing tissue effects

Here I look explicitly at whether there is a significant association between overall expression patterns and tissue site.

Tissue effect
	Df	SumsOfSqs	MeanSqs	F.Model	R2	Pr(>F)
Tissue.location	2	0.0997791	0.0498896	25.51449	0.3924431	0.001
Residuals	79	0.1544721	0.0019553	NA	0.6075569	NA
Total	81	0.2542512	NA	NA	1.0000000	NA

PCA at each tissue

Here I look at the principle components analysis at each tissue location separately. This will help to detect if there are any differences by disease status that are somewhat masked by the huge amount of variation explained by tissue location.

There is not a clear clustering by group in terms of overall gene expression patterns. Nevertheless we can test for association using a PERMANOVA.

PERMANOVA for disease associations at each tissue

Below I test whether there is a difference in overall expression profiles by disease status. This is done in each tissue separately.

Ileum
	Df	SumsOfSqs	MeanSqs	F.Model	R2	Pr(>F)
Disease	2	0.0059934	0.0029967	1.614911	0.1186134	0.051
Residuals	24	0.0445357	0.0018557	NA	0.8813866	NA
Total	26	0.0505292	NA	NA	1.0000000	NA

Caecum
	Df	SumsOfSqs	MeanSqs	F.Model	R2	Pr(>F)
Disease	2	0.0032406	0.0016203	1.061678	0.0845172	0.36
Residuals	23	0.0351015	0.0015262	NA	0.9154828	NA
Total	25	0.0383421	NA	NA	1.0000000	NA

Rectum
	Df	SumsOfSqs	MeanSqs	F.Model	R2	Pr(>F)
Disease	2	0.0041217	0.0020609	0.8715505	0.0628301	0.653
Residuals	26	0.0614791	0.0023646	NA	0.9371699	NA
Total	28	0.0656008	NA	NA	1.0000000	NA

There look to be very subtle overall differences in overall gene expression profiles in terms of disease status. The strongest effects are in the ileum as noted by the highest R^2.

Differential expression analysis by group (disease)

Below is the results of differential expression analysis using DESeq2. No covariates are taken into account (numbers are small). Initially I perform an overall test (LRT) to see which genes are differentially expressed between any group.

Number of significantly differentially expressed genes at each site.

nsig.caecum	nsig.ileum	nsig.rectum
17	111	27

## png 
##   2

Heatmap of changes

Below is a heatmap of all of the observed changes (union at each tissue site). This is to try to get a handle on whether there are any patterns there. I am also including the metadata to see if there are any obvious associations with factors such as inflammation.

Differential expression analysis

In the ileum, the PSC/UC and UC look fairly similar. The caecum looks to be more affected in PSC and the rectum more in UC. These patterns are more or less what we would have expected. As opposed to the tissue-defining clusters that we found there is no obvious clustering here and so pairwise comparisons are potentially going to identify more robust differences.

DESeq2 analysis at each site

Below I run the analysis looking at each tissue separately and each pairwise comparison. The table summarises the results of these comparisons. We do not have enough samples to assess the impact of inflammation - in fact the vast majority of samples are not inflamed.

Table of results of differential expression

Below are tables that summarise the differentially expressed genes between each disease group at each tissue site.

Ileum: PSC/UC vs Healthy
Var1	Freq
Down	45
Up	151

Ileum: UC vs Healthy
Var1	Freq
Down	302
Up	427

Ileum: PSC/UC vs UC
Var1	Freq
Down	1
Up	2

Caecum: PSC/UC vs Healthy
Var1	Freq
Down	174
Up	46

Caecum: UC vs Healthy
Var1	Freq
Down	2
Up	1

Caecum: PSC/UC vs UC
Var1	Freq
Down	14
Up	2

Rectum: PSC/UC vs Healthy
Var1	Freq
Up	1

Rectum: UC vs Healthy
Var1	Freq
Down	7
Up	30

Rectum: PSC/UC vs UC
Freq

Heatmap the significant differences

Below I draw a heatmap to show the union of significant differences between any pairwise comparison. I also output a venn diagram to show the overlaps between the tissue sites.

## png 
##   2

Ileum heatmap

Caecum heatmap

Rectum heatmap

Correlating effect sizes

In the following analysis I look at correlating fold changes between the two pairwise comparisons (i.e. each disease vs. healthy). This will reveal differences between the two disease conditions. It is exprected that if they are truly distinct in transcriptomic signatures that they will not be highly correlated.

	ileum.cor	ileum.p	caecum.cor	caecum.p	rectum.cor	rectum.p
cor	0.6748169	0	0.5055982	0	0.420959	0

## png 
##   2

## png 
##   2

## png 
##   2

These are certainly significantly correlated. The ileum shows some interesting patterns - especially for a handful of genes that look as if they are downregulated in UC but up-regulate in PSC. Should see what these are.

Markers of inflammation

Very few of our samples have histological eveidence of inflammation. Nevertheless we are interested in whether there are any molecular (i.e. gene expression) markers of inflammation that might hint at any subclinical inflammation that might be present in the differnetial expression results. To test this I look at a few marker genes that are known to be robustly associated with ulcerative colitis namely, TNF, IL17A, IL12A, IL23A, NOS2 and S100A8. To look at this, I take the union of significantly differentially expressed genes from PSC/UC vs. Healthy and UC vs. Healthy and see whether these genes are differentially expressed.

It’s kind of interesting because the inflammatory markers that I have chosen are only increased in the ileum - this is supposed to be a relatively uninvolved site. There may be some molecular inflammation that does not manifest as overt inflammation. By eye there are no obvious associations with any of the measured covariates like drugs or histology (Nancy here is the Nancy index at the tissue site sampled).

Enrichment of tissue clusters

I initially had a hypothesis that the reason for not seeing very strong overlap in disease associations between tissue sites was because tissue-defining gene expression signatures were the things that were altered in disease states. Here I explore this hypothesis by looking at the enrichment of previously defined tissue clusters amongst genes that are significantly differentially expressed in either PSC/UC or UC.

code	scount	stotal	spercent	bcount	btotal	bpercent	ratio	pvalue	pover	punder	goid	category	description	fdr	Type
-	33	131	25.19	4148	12995	31.92	0.79	0.0564600	0.9626300	0.0564600	Cluster 1	Tissue.clusters	Cluster 1	0.070575	Ileum PSC vs. Healthy
+	41	131	31.30	3081	12995	23.71	1.32	0.0282480	0.0282480	0.9822000	Cluster 2	Tissue.clusters	Cluster 2	0.047080	Ileum PSC vs. Healthy
-	28	131	21.37	2849	12995	21.92	0.97	0.4891600	0.5950200	0.4891600	Cluster 3	Tissue.clusters	Cluster 3	0.489163	Ileum PSC vs. Healthy
+	25	131	19.08	1614	12995	12.42	1.54	0.0182420	0.0182420	0.9900300	Cluster 4	Tissue.clusters	Cluster 4	0.045605	Ileum PSC vs. Healthy
-	4	131	3.05	1364	12995	10.50	0.29	0.0013872	0.9996600	0.0013872	Cluster 5	Tissue.clusters	Cluster 5	0.006936	Ileum PSC vs. Healthy
-	29	169	17.16	4148	12995	31.92	0.54	0.0000097	1.0000000	0.0000097	Cluster 1	Tissue.clusters	Cluster 1	0.000012	Caecum PSC vs. Healthy
-	10	169	5.92	3081	12995	23.71	0.25	0.0000000	1.0000000	0.0000000	Cluster 2	Tissue.clusters	Cluster 2	0.000000	Caecum PSC vs. Healthy
-	15	169	8.88	2849	12995	21.92	0.40	0.0000052	1.0000000	0.0000052	Cluster 3	Tissue.clusters	Cluster 3	0.000009	Caecum PSC vs. Healthy
+	104	169	61.54	1614	12995	12.42	4.95	0.0000000	0.0000000	1.0000000	Cluster 4	Tissue.clusters	Cluster 4	0.000000	Caecum PSC vs. Healthy
-	11	169	6.51	1364	12995	10.50	0.62	0.0509010	0.9729300	0.0509010	Cluster 5	Tissue.clusters	Cluster 5	0.050901	Caecum PSC vs. Healthy
+	1	1	100.00	4148	12995	31.92	3.13	0.3192000	0.3192000	1.0000000	Cluster 1	Tissue.clusters	Cluster 1	0.895037	Rectum PSC vs. Healthy
-	0	1	0.00	3081	12995	23.71	0.00	0.7629100	1.0000000	0.7629100	Cluster 2	Tissue.clusters	Cluster 2	0.895037	Rectum PSC vs. Healthy
-	0	1	0.00	2849	12995	21.92	0.00	0.7807600	1.0000000	0.7807600	Cluster 3	Tissue.clusters	Cluster 3	0.895037	Rectum PSC vs. Healthy
-	0	1	0.00	1614	12995	12.42	0.00	0.8758000	1.0000000	0.8758000	Cluster 4	Tissue.clusters	Cluster 4	0.895037	Rectum PSC vs. Healthy
-	0	1	0.00	1364	12995	10.50	0.00	0.8950400	1.0000000	0.8950400	Cluster 5	Tissue.clusters	Cluster 5	0.895037	Rectum PSC vs. Healthy
-	104	454	22.91	4148	12995	31.92	0.72	0.0000104	0.9999900	0.0000104	Cluster 1	Tissue.clusters	Cluster 1	0.000026	Ileum UC vs. Healthy
+	184	454	40.53	3081	12995	23.71	1.71	0.0000000	0.0000000	1.0000000	Cluster 2	Tissue.clusters	Cluster 2	0.000000	Ileum UC vs. Healthy
-	92	454	20.26	2849	12995	21.92	0.92	0.2093100	0.8228800	0.2093100	Cluster 3	Tissue.clusters	Cluster 3	0.242122	Ileum UC vs. Healthy
-	51	454	11.23	1614	12995	12.42	0.90	0.2421200	0.8018800	0.2421200	Cluster 4	Tissue.clusters	Cluster 4	0.242122	Ileum UC vs. Healthy
-	26	454	5.73	1364	12995	10.50	0.55	0.0002020	0.9999000	0.0002020	Cluster 5	Tissue.clusters	Cluster 5	0.000337	Ileum UC vs. Healthy
-	4	20	20.00	4148	12995	31.92	0.63	0.1845300	0.9225500	0.1845300	Cluster 1	Tissue.clusters	Cluster 1	0.184530	Rectum UC vs. Healthy
-	0	20	0.00	3081	12995	23.71	0.00	0.0044410	1.0000000	0.0044410	Cluster 2	Tissue.clusters	Cluster 2	0.011102	Rectum UC vs. Healthy
+	8	20	40.00	2849	12995	21.92	1.82	0.0525450	0.0525450	0.9819000	Cluster 3	Tissue.clusters	Cluster 3	0.087575	Rectum UC vs. Healthy
+	8	20	40.00	1614	12995	12.42	3.22	0.0017596	0.0017596	0.9996800	Cluster 4	Tissue.clusters	Cluster 4	0.008798	Rectum UC vs. Healthy
-	0	20	0.00	1364	12995	10.50	0.00	0.1086600	1.0000000	0.1086600	Cluster 5	Tissue.clusters	Cluster 5	0.135825	Rectum UC vs. Healthy

It doesn’t look exactly how I was expecting. In fact, particularly in the ileum there is an enrichment of cluster 2 genes (proliferation cluster) amongst differentially expressed genes. Below I explore in a little more detail what these are and how they relate to tissue-defining clusters.

Cluster 2 genes altered in PSC/UC

Interestingly it is not particularly genes that differentiate the ileum from other tissues that are altered in PSC/UC. Indeed, there is an enrichment of cluster 2 genes. Below I explore what these genes are.

Heatmap of everything. Will see how this looks in terms of adding into the manuscript.

Cluster 4 genes in PSC/UC

There seems to be an increase in proliferative cells in PSC/UC ileum. I next investigate whether this is linked to B-cells i.e. those that are expressing antibody (BCR) which are cluster 4 genes. This would imply that there is an increase in proliferating B-cells in disease ileum vs. healthy.

Cluster 4 genes in Caecum PSC/UC

Cluster 4 genes are enriched in PSC/UC associated genes. Here I explore what these are.

FOXP3 is pretty interesting especially in PSC. This is because it seems to be up in the ileum and down in the caecum. This may suggest that these two tissues are linked by Tregs. If this was the case then we would expect to see a negative correlation between FOXP3 expression in the ileum vs. the caecum.

There is no real correlation there so it is unlikely that the effects in one tissue are affecting the other.

Correlation between IL7R and FOXP3

It is of interest that FOXP3 and IL7R are both downregulated in PSC/UC. The relationship between these two genes is somewhat hard to interpret - IL7 is important for Treg development in the thymus but it seems that in inflammation IL7R expression on a subset of T-cells (not Tregs) represents a pathogenic class of T cells. There is also evidence that activated Tregs express high levels of IL7R. Below I look at the correlation beween IL7R and FOXP3 expression within patients with PSC/UC. If co-expressed it would suggest that these genes are expressed from the same cell types.

These data are interesting as there is a definite positive correlation between FOXP3 and IL7R. I did a quick check for whether there was a significant difference between correlation coefficients PSC/UC and UC and there isn’t (data not shown here). It is likely that the numbers are so small that we don’t get a robust correlation. Therefore I would like to use more data - to this end I am going to try to remove the effect of disease from the expression measurements and do the correlations on all of the data at the same time. This should remove spurous correlation that could occur due to differences between groups but retain power to get a robust correlation. I do this using the limma function removeBatchEffect().

This appears to represent the average correlation which is what would be expected. The increased numbers means that the correlation is significant - it comes down to what to present in the manuscript. There is a balance between something that could end up being confusing due to removal of batch effects or confusing due to differences in correlations between groups. I think I will go for the former as I think it best represents the data.

Given the correlation between IL7R and FOXP3 and evidence that activated Tregs express high levels of IL7R, I thought it would be worth looking at the correlation between these genes and the activation markers, ICOS, CTLA4 and CD103 (ITGAE).

I think that this is pretty convincing evidence that we are looking at activated T cells here. Reductions of FOXP3 and IL7R and likely due to reductions in activated T cells - indeed ICOS is slightly (although not significantly donwregulated in psc). I next explore how to present these data potentially in a single figure rather than multiple scatterplots.

```{all.cor.activation, echo=FALSE, message=FALSE, fig.height=5, fig.width=5}

cors.activation <- corr.test(df.compare.il7r.foxp3[,c(1:7)]) cors.activation.r <- cors.activation\(r cors.activation.p <- cors.activation\)p

cor.colors <- colorRampPalette(c(“white”, blues9[6]))(3) p.cors <- ggcorrplot(cors.activation.r, hc.order = TRUE, type = “upper”, lab=TRUE, colors = cor.colors, p.mat=cors.activation.p) p.cors ggsave(“export.dir/cors_activation.pdf”, plot=p.cors + theme(text=element_text(size=6)), height=3, width=3.5)


I'm not convinced that I like just presenting the correlations and so will see if I can fit the scatterplots in the manuscript. 


#### Cluster 4 genes in Rectum UC

<img src="disease_differential_expression_files/figure-html/rectum.uc.cluster4.genes-1.png" width="1440"  />


#### Cluster 4 genes in Rectum UC

<img src="disease_differential_expression_files/figure-html/rectum.uc.cluster3.genes-1.png" width="1440"  />

### PSC-specific changes

Although globally PSC/UC and UC look quite similar i.e. correlated effect sizes, we did still call a handful of genes as differentially abundant between PSC/UC and UC. Below are the plots of genes called as significantly differentially expressed in each tissue site. 

#### Ileum PSC/UC vs. UC

<img src="disease_differential_expression_files/figure-html/psc.specific.ileum-1.png" width="1440"  />

#### Caecum PSC/UC vs. UC

<img src="disease_differential_expression_files/figure-html/psc.specific.caecum-1.png" width="1440"  />

#### Rectum PSC/UC vs. UC

No genes were significantly differentially expressed between PSC/UC and UC in the rectum.

### GGT1 expression

Manually looking at these genes suggests that these are predominantly lowly expressed and only seen as different in a handful of patients, especially in the ileum. However, one gene particularly stands out and that is GGT1 - it is significantly lower in expression in both the caecum and the ileum.

This gene is potentially very interesting as it regulates glutathione scavenging. It is expressed at the cell membrane and transfers groups from glutahione to cysteine. This then allows uptake of the molecule into cells to be converted back into glutathione that is useful for mopping up reactive oxygen etc. This therefore leads to an hypothesis about its role in inflammation-driven cancer - in the presence of inflammation where there is an increase in reactive oxygen, PSC patients are less able to protect against reactive oxygen-driven DNA-damgage. This may then lead to an increase in the number of somatic, cancer-driving mutations. This is only likely the case where there is inflammation - hence why the ileum may be spared from cancer in these patients.

#### GGT1 and liver function tests

We have data on the outcomes of liver function tests i.e. bilirubin, Alkaline Phosphatase (ALP) and Alanine aminotransferase (ALT). Unfortunately we don't have GGT analysis although I am hoping that we can use these other biochemical measurements as markers of the extent of cholestasis. 

#### Biochemical analysis between groups

We have some data for all groups (although some are missing data on liver function tests). Here I just plot the levels that we observe for each individual.

<img src="disease_differential_expression_files/figure-html/biochemistry-1.png" width="960"  />

### Correlations with GGT1

I am interested in whether GGT1 expression is associated with bile acid transport - in this way the hypothesis is that it would be negatively coreelated with biochemical markers of cholestasis. Below I look at the correlation between GGT1 and biochemical markers in each tissue site.

#### Bilirubin

<img src="disease_differential_expression_files/figure-html/cor.bil.with.ggt1-1.png" width="1440"  />

#### ALT

<img src="disease_differential_expression_files/figure-html/cor.alt.with.ggt1-1.png" width="1440"  />

#### ALP

<img src="disease_differential_expression_files/figure-html/cor.alp.with.ggt1-1.png" width="1440"  />

There isn't a clear relationship between GGT1 and biochemical markers of cholestasis, although fairly strong negative cor between GGT1 and bilirubin in the ileum. There are a few strong positive associations for ALT and ALP in the rectum. These are however quite dirven by an outlier. I think with such few samples it is very difficult to draw any conclusions from these data.



### CIBERSORT analysis

It was suggested that I run CIBERSORT in order to determine if there were any differences in cell populations between the disease states. This was particularly suggested by Beth to see if there was a reduction in Tregs in the Caecum in PSC/UC. I have used single cell expression data from the [human gut cell atlas](https://www.gutcellatlas.org) in order to generate a custom signatures matrix. This was then used for the deconvolution of cell fractions in our RNA-seq data. I ran this using the web application with the following parameters:

>Running CIBERSORTxFractions...
>[Options] mixture: files/nicholas.ilott@kennedy.ox.ac.uk/cpm_cyber.tsv
>[Options] sigmatrix: files/nicholas.ilott@kennedy.ox.ac.uk/CIBERSORTx_Job12_immune_cells_inferred_phenoclasses.CIBERSORTx_Job12_immune_cells_inferred_refsample.bm.K999.txt
>[Options] perm: 1000
>[Options] verbose: TRUE
>[Options] QN: FALSE
>[Options] outdir: files/nicholas.ilott@kennedy.ox.ac.uk/results/
>[Options] label: Job13
>=============CIBERSORTx Settings===============
>Mixture file: files/nicholas.ilott@kennedy.ox.ac.uk/cpm_cyber.tsv 
>Signature matrix file: files/nicholas.ilott@kennedy.ox.ac.uk/CIBERSORTx_Job12_immune_cells_inferred_phenoclasses.CIBERSORTx_Job12_immune_cells_inferred_refsample.bm.K999.txt 
>Number of permutations set to: 1000 
>Enable verbose output



#### CIBERSORT in the ileum

<img src="disease_differential_expression_files/figure-html/ileum.ciber-1.png" width="2400"  />

#### CIBERSORT in the caecum

<img src="disease_differential_expression_files/figure-html/caecum.ciber-1.png" width="2400"  />

#### CIBERSORT in the rectum

<img src="disease_differential_expression_files/figure-html/rectum.ciber-1.png" width="2400"  />

There is nothing major going on in terms of predicted cell frequencies. This is likely due to the complexity of the regulation as assessed by bulk RNA-seq. 


### Visualisation in heatmap

It looks like the ComplexHeatmap package might come to the rescue and be able to get the numerous genes, clusters and conditions plotted in a single heatmap so that the data do not have to be displayed as boxplots. I explore below whether this is the case:

png

2 null device

1


### Session Info

R version 3.6.1 (2019-07-05)

Platform: x86_64-conda_cos6-linux-gnu (64-bit)

Running under: CentOS release 6.7 (Final)

Matrix products: default

BLAS/LAPACK: /gfs/devel/nilott/cgat-24102019/conda-install/envs/cgat-flow/lib/libopenblasp-r0.3.9.so

locale:

1 LC_CTYPE=en_GB.UTF-8 LC_NUMERIC=C

[3] LC_TIME=en_GB.UTF-8 LC_COLLATE=en_GB.UTF-8

[5] LC_MONETARY=en_GB.UTF-8 LC_MESSAGES=en_GB.UTF-8

[7] LC_PAPER=en_GB.UTF-8 LC_NAME=C

[9] LC_ADDRESS=C LC_TELEPHONE=C

[11] LC_MEASUREMENT=en_GB.UTF-8 LC_IDENTIFICATION=C

attached base packages:

1 grid parallel stats4 stats graphics grDevices utils

[8] datasets methods base

other attached packages:

1 VennDiagram_1.6.20 futile.logger_1.4.3

[3] reshape_0.8.8 gtools_3.8.2

[5] data.table_1.12.8 vegan_2.5-6

[7] lattice_0.20-41 permute_0.9-5

[9] psych_1.9.12.31 ComplexHeatmap_2.2.0

[11] ggcorrplot_0.1.3 ggpubr_0.2.5

[13] magrittr_1.5 dplyr_0.8.0.1

[15] ggplot2_3.2.1 pheatmap_1.0.12

[17] RSQLite_2.1.2 vsn_3.54.0

[19] DESeq2_1.26.0 SummarizedExperiment_1.16.0

[21] DelayedArray_0.12.0 BiocParallel_1.20.0

[23] matrixStats_0.56.0 Biobase_2.46.0

[25] GenomicRanges_1.38.0 GenomeInfoDb_1.22.0

[27] IRanges_2.20.0 S4Vectors_0.24.0

[29] BiocGenerics_0.32.0 knitr_1.25

[31] gridExtra_2.3 gplots_3.0.3

[33] RColorBrewer_1.1-2

loaded via a namespace (and not attached):

1 colorspace_1.4-1 ggsignif_0.6.0 rjson_0.2.20

[4] ellipsis_0.3.0 circlize_0.4.10 htmlTable_1.13.3

[7] XVector_0.26.0 GlobalOptions_0.1.2 base64enc_0.1-3

[10] clue_0.3-57 rstudioapi_0.11 farver_2.0.3

[13] affyio_1.56.0 bit64_0.9-7 AnnotationDbi_1.48.0

[16] splines_3.6.1 mnormt_1.5-6 geneplotter_1.64.0

[19] Formula_1.2-3 annotate_1.64.0 cluster_2.1.0

[22] png_0.1-7 BiocManager_1.30.10 compiler_3.6.1

[25] backports_1.1.6 assertthat_0.2.1 Matrix_1.2-18

[28] lazyeval_0.2.2 limma_3.42.0 formatR_1.7

[31] acepack_1.4.1 htmltools_0.4.0 tools_3.6.1

[34] gtable_0.3.0 glue_1.4.0 GenomeInfoDbData_1.2.1

[37] affy_1.64.0 Rcpp_1.0.4.6 vctrs_0.2.4

[40] gdata_2.18.0 preprocessCore_1.48.0 nlme_3.1-147

[43] xfun_0.13 stringr_1.4.0 lifecycle_0.2.0

[46] XML_3.98-1.20 MASS_7.3-51.4 zlibbioc_1.32.0

[49] scales_1.1.0 lambda.r_1.2.4 memoise_1.1.0

[52] rpart_4.1-15 latticeExtra_0.6-29 stringi_1.4.3

[55] highr_0.8 genefilter_1.68.0 checkmate_2.0.0

[58] caTools_1.18.0 shape_1.4.4 rlang_0.4.5

[61] pkgconfig_2.0.3 bitops_1.0-6 evaluate_0.14

[64] purrr_0.3.4 labeling_0.3 htmlwidgets_1.5.1

[67] bit_1.1-15.2 tidyselect_1.0.0 plyr_1.8.6

[70] R6_2.4.1 Hmisc_4.2-0 DBI_1.0.0

[73] mgcv_1.8-31 pillar_1.4.3 foreign_0.8-76

[76] withr_2.1.2 survival_3.1-12 RCurl_1.95-4.12

[79] nnet_7.3-13 tibble_3.0.1 crayon_1.3.4

[82] futile.options_1.0.1 KernSmooth_2.23-16 rmarkdown_1.16

[85] jpeg_0.1-8.1 GetoptLong_1.0.0 locfit_1.5-9.4

[88] blob_1.2.1 digest_0.6.25 xtable_1.8-4

[91] munsell_0.5.0

```